Netgear XAVB5001 (Atheros 500Mbps) vs Zyxel PLA4205 (Atheros 500Mbps) vs Comtrend "BT Vision Powerline" (Marvell/DS2 200Mbps) vs Solwise Net-PL-1000M (Gigle 1000Mbps)
I suspect there might be only an audience of one here, so instead of all the plots, let me cut a long story as short I can.
Performance
In this group, the Netgears perform the best. Throughput is fairly steady: 90Mbps on the same outlet (I saw no overloading issues on any of these adapters, same outlet was always the top performer), 24Mbps at the worst location and anything inbetween everywhere else. Putting the plugs on the end of surge protectors and mains filters did indeed reduce throughput by around 20-30% so really must be avoided. This was tested in a large house (1500 sq ft per floor), with UK 230V ring mains.
Due primarily due to the latency of the powerline adapters, the Ultra High Performance script of IxChariot corresponds best (in fact, almost exactly) to a simple Windows file transfer test and these are how I obtained the results above.
The throughput is not predictable based on distance from circuit breaker or even passing through circuit breakers at all. One of the best results was on a different mains ring, and both at the furthest point from the circuit breakers, so go figure. Location drop-offs in throughput were consistent between different plugs and technologies though, so something is happening and predictable somewhere, I just dont know what.
The Zyxel performs significantly worse than the Netgear at a lot of locations, and achieves a lower reported link rate when this happens, eg at one location I saw 60Mbps vs 82Mbps real throughput and 130Mbps vs 200Mbps link rate. The link rate for all these devices is constantly being re-negotiated (and is very different between Tx and Rx a lot of the time), especially at plug-in: I tried multiple resets and waits and data transfers before testing and re-testing the Zyxel, this is a certain result. Pairing up one Zyxel to one Netgear gave Zyxel-only-pair results. The firmware reported for the Zyxel is from March 2011, from the Netgear November 2010.
The Comtrend DS2 has steady throughput but is about 1/2 to 1/3rd the throughput of the Netgear at all locations.
The Solwise gigabit has regular bandwidth drops, its mountain valley. It can get a good average throughput at the same outlet though, rivalling the 500Mbps adapters, but from the next room onwards is in DS2 200Mbps territory. This adapter has a constant steady response time of just under 2ms, whereas the 500Mbps adapters have spikes of around 12ms every 15 seconds or so.
Design, features, reliability
The first Netgears I got were faulty and had low throughput, the replacements seem fine. The Zyxel plug connected to the router sometimes had trouble waking up from standby, and a couple of times they just threw a fit, stopped working and needed to be replugged in. The Netgears run a lot hotter to the touch than any other device here.
The Netgears look the classiest and dont have blindingly bright and blinking LEDs unlike the others (they can even also be turned off completely in the utility, this is neat thinking for a home feature). The Comtrend ethernet cable comes out the top which can be practical behind a desk, but looks unsightly out on show. The Zyxels look good, the Solwise have a cheap, bright, white plastic finish.
The 500Mbps adapters here have an auto power-saving mode by monitoring the ethernet connection: the Netgear goes off after 10 mins, the Zyxel after 1 min, they wake up on reconnection in just a few seconds. In standby, they both emit an audible high-frequency noise but its not too loud whereas the Solwise gives out a much louder high-frequency noise all the time, if its close to where you sit, you may soon get a headache like I did. The Netgears can QoS by plug priority as well as ports, the Zyxels just by plug priority.
They all function just like a cable-replacement, so you can simply plug in any £15 switch if you want to connect one plug to many end-devices. They really are as easy to setup as the "2 seconds to push a button" claims. You can run independent networks on the same mains by giving the homplugs different network names. The Homeplugs all worked and paired with each other fine, though the Zyxel utility has no facility to enter an encryption key manually so I had to use hardware button to set an unknown random key for all adapters.
Video Streaming
I must mention a side-note on video streaming: your plumbing is only as good as your taps: the end devices, OS and software are critical. In Windows, Quicktime and Windows Media Player should not be used to assess video streaming performance, as they do not perform in this role properly and need far greater bandwidth to play back files than should be necessary. With a buffer to blu-ray spec, a max data rate of 54Mbps is all that should be required. With other players, VLC being my preferred choice (J River Media Center and XBMC performed just as well too), you get much better results. Via DLNA to PS3, results were much better than Quicktime, but quite a bit worse than software players on Windows.
So this is good news: SNB's supposed torture test of the MegaMind 1080p trailer now becomes a breeze. A steady 26Mbps is adequate. I bumped up the buffer to 1 sec in VLC, and 802.11g handled it fine. After bumping it up to 2 seconds, I was going from one wireless laptop to another wireless laptop the other side of the house on 802.11g without hiccup!
But here's the bad news: the MegaMind trailer is not a good test to give a clean bill of health for HD video. A slightly tougher test from the Apple trailers is Mission Impossible Ghost Protocol: if you can play this you can play a lot of HD videos fine. But there are two much tougher tests: the Melancholia trailer, with a higher and more sustained "max average" data rate. And also an action sequence backed-up direct from Casino Royale blu-ray, with similarly high "max average" as the Melancholia trailer but also a much higher average bitrate as you would expect. If you can stream these two tests, Melancholia trailer and Casino Royale blu-ray, without flaw, then you can be pretty sure you can play back almost all commercially distributed HD video. With the Netgears, and VLC on PC with default settings for playback, I managed this at most locations. With DLNA to PS3, I did not manage it at most locations. With a dedicated streaming hardware device like a Popcorn Hour, who knows...
In sum: the 500Mbps Atheros adapters perform best at all locations. But they are far from equal, so you can't just buy going on chipset. If video streaming, choose your playback setup carefully.
Expect a Netgear vs Devolo vs Billion vs "Cheap Adapter" shootout very soon... !